xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-03-25 19:37:05 +08:00
Code Issues Packages Projects Releases Wiki Activity

align with_hf_config

Browse Source 
Signed-off-by: Xingyu Liu <charlotteliu12x@gmail.com>
This commit is contained in:
Xingyu Liu 2025-12-05 09:59:59 -08:00
parent ec81168330
commit 9b19e3b94f
2 changed files with 9 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
vllm/config/model.py
View File

@ -508,12 +508,7 @@ class ModelConfig:
self.hf_image_processor_config = get_hf_image_processor_config(
self.model, hf_token=self.hf_token, revision=self.revision
)
self.model_arch_config = None
convertor_cls = MODEL_ARCH_CONFIG_CONVERTORS.get(
hf_config.model_type, ModelArchConfigConvertorBase
)
convertor = convertor_cls(hf_config)
self.model_arch_config = convertor.convert(self.model, self.revision)
self.model_arch_config = self.get_model_arch_config()
architectures = self.architectures
registry = self.registry
@ -717,6 +712,13 @@ class ModelConfig:
self._verify_cuda_graph()
self._verify_bnb_config()
def get_model_arch_config(self) -> ModelArchitectureConfig:
convertor_cls = MODEL_ARCH_CONFIG_CONVERTORS.get(
self.hf_config.model_type, ModelArchConfigConvertorBase
)
convertor = convertor_cls(self.hf_config)
return convertor.convert(self.model, self.revision)
@field_validator("tokenizer_mode", mode="after")
def _lowercase_tokenizer_mode(cls, tokenizer_mode: str) -> str:
return tokenizer_mode.lower()

1
vllm/config/vllm.py
View File

@ -421,6 +421,7 @@ class VllmConfig:
model_config = copy.deepcopy(self.model_config)
model_config.hf_config = hf_config
model_config.model_arch_config = model_config.get_model_arch_config()
return replace(self, model_config=model_config)